Analytics X Prize - Predicting Homicides in Philadelphia | Hacker News Reader

Analytics X Prize - Predicting Homicides in Philadelphia

analyticsx.com

20 pointstshtf13 comments

To precisely state what others have implied - the prize was deliberately set low so that there would be no incentive to murder someone in order to win the prize. The prize is "symbolic" rather than "incentive."

If the prize was $1mm, then someone out there might have been inspired to murder people in particular areas after predicting they would be killed in that area.

Seriously $100 to do statistical work? The site looks very nicely laid out, but I have to wonder what the purpose of this competition is.

My first inclination with such a low prize amount is to be a little suspicious of the motives behind this effort. For instance, is some local political figure wanting to do something that might be viewed as controversial and this site is a CYA move before taking that action?

What I mean is will this site be used by some politician to set himself up to make the statement in public "We tried to get X done with this website campaign, paid millions of the taxpayers money for web development, even offered a cash prize (without saying how much the prize was) but the results were trivial so now we have no other option than to do Y (whatever controversial thing the politician’s contributors wanted done in the first place).".

To the OP: if this is your site and you are seriously looking to promote it and attract contestants you might want to think about bumping the $100 grand prize up to at least $1,000. It seems to me the $100 will not be enough (after taxes) for people to take the effort to get involved. Maybe seek monetary sponsorship from the city for the prize amount.

This doesn't seem to have any relation to the "X Prize" as run by http://www.xprize.org/

you can tell by the prize amounts...

Any system which has human psychology in the loop is going to be difficult to predict. I'll be impressed if these kinds of phenomena can be predicted though. We already know that crime is more likely in certain neighborhoods, but whether you can make more specific predictions than that I don't know (almost getting into precrime territory).

"A Grand Prize ($100) is awarded at the end of the calendar year to the best performing submission."

This is probably a good thing. A large prize might create some rather perverse incentives for, uhh, direct intervention in the data set.

Blech. This is not an easy problem and that's all they offer? <snarky>Is that the dollar value Philly puts on a life?</snarky>

Not understanding the downvotes. Solving this problem would easily save hundreds of thousands of dollars, if not millions, in law enforcement and litigation costs. To put a paltry $100 reward for _preventing murder_ is, frankly, pathetic.

The prize isn't for preventing murder; it's for predicting murder. There are significant practical differences, particularly with respect to the incentives created.

In fact, if your submission helps prevent murders, you would tend to lose the contest.

Well, predicting murders would cause a re-allocation of the police department's resources, arguably preventing some of those murders.

Being able to predict where/when murder happens leads to a more efficient and effective distribution of law enforcement. Other cities have done this with non-violent crime/traffic violations/etc., and crime has gone down. So prediction = prevention.

Reply on news.ycombinator.com